Erythropoietin (EPO) is a 30.4 kilodalton (kDa) glycoprotein hormone that promotes the proliferation of erythroid progenitor cells and supports their differentiation into mature erythrocytes (see, for example, Krantz, Blood, 77:419-434, 1991). EPO is produced in the adult kidney and the fetal liver. In adults, EPO is produced primarily in kidney cells in response to hypoxia or anemia and circulates in the bloodstream. EPO targets the 66 kDa specific receptor (EPO-Rc) found almost exclusively on the surface of erythroid progenitor cells present in bone marrow. Upon binding EPO, the receptor is activated and undergoes homodimerization, followed by tyrosine phosphorylation. Subsequently, a series of intracellular signal transduction events take place, leading to the increase of the number of the progenitor cells and their maturation into erythrocytes (see, for example, Lodish et al., Cold Spring Harbor Symp. Quant. Biol., 60:93-104, 1995).
Recombinant human EPO (rHuEPO) is widely used in the treatment of patients with chronic anemia due to renal diseases at both end-stage and pre-dialysis phases. Administration of EPO has also been successful to treat anemia in patients caused by cancer chemotherapy, rheumatoid arthritis, AZT treatment for HIV infection and myelodysplastic syndrome. No direct toxic effect of treatment has been reported and the benefits of blood transfusion could be achieved without the transfusion.
The concentration of EPO in normal human serum varies approximately from 0.01 to 0.03 units/ml. Supplemental EPO is a desirable treatment in cases of renal failure with decreased EPO production. The half-life for the serum clearance of intravenous (i.v.) rHuEPO is approximately 4 to 13 h. The peak serum concentration for subcutaneous (s.c.) rHuEPO occurs in 5 to 24 h after injection with an elimination half-life of 17 h. The s.c. administration route can therefore lead to much longer retention in the blood than i.v. administration of the same dose. The mechanism responsible for clearing EPO from the serum remains unclear. In animal experiments, less than 5% is excreted by the kidney. The liver, which rapidly removes asialated EPO, has not been shown to play a significant role in clearing EPO (see, for example, Fried, Annu. Rev. Nutr., 15:353-377, 1995).
Immunoglobulins of IgG class are among the most abundant proteins in human blood. Their circulation half-lives can reach as long as 21 days. Fusion proteins have been reported to combine the Fc regions of IgG with the domains of another protein, such as various cytokines and soluble receptors (see, for example, Capon et al., Nature, 337:525-531, 1989; Chamow et al., Trends Biotechnol., 14:52-60, 1996); U.S. Pat. Nos. 5,116,964 and 5,541,087). The prototype fusion protein is a homodimeric protein linked through cysteine residues in the hinge region of IgG Fc, resulting in a molecule similar to an IgG molecule without the CHl domains and light chains. Due to the structural homology, Fc fusion proteins exhibit in vivo pharmacokinetic profile comparable to that of human IgG with a similar isotype. This approach has been applied to several therapeutically important cytokines, such as IL-2 and IFN-α2a, and soluble receptors, such as TNF-Rc and IL-5-Rc (see, for example, U.S. Pat. Nos. 5,349,053 and 6,224,867). To extend the circulating half-life of EPO and/or to increase its biological activity, it is desirable to make fusion proteins containing EPO linked to the Fc portion of the human IgG protein as disclosed or described in this invention.
In most of the reported Fc fusion protein molecules, a hinge region serves as a spacer between the Fc region and the cytokine or soluble receptor at the amino-terminus, allowing these two parts of the molecule to function separately (see, for example, Ashkenazi et al., Current Opinion in Immunology, 9:195-200, 1997). Relative to the EPO monomer, a fusion protein consisting of two complete EPO domains separated by a 3- to 7-amino acid peptide linker exhibited reduced activity (Qiu et al., J. Biol. Chem., 273:11173-11176, 1998). However, when the peptide linker between the two EPO domains was 17 amino acids in length, the dimeric EPO molecule exhibited considerably enhanced in vitro and in vivo activities. The enhanced activity has been shown to be due to an increased in vitro activity coupled with a different pharmacokinetic profile in mice (see, for example, Sytkowski et al., J. Biol. Chem., 274:24773-24778, 1999; U.S. Pat. No. 6,187,564 ). A human EPO fusion protein with an appropriate peptide linker between the HuEPO and Fc moieties (HuEPO-L-Fc) is more active than rHuEPO, with in vitro activity at least 2-fold as that of rHuEPO on a molar basis. It is discovered according to this invention that an added peptide linker present between HuEPO and a human IgG Fc variant enhances the in vitro biological activity of the HuEPO-L-Fc molecule in two ways: (1) keeping the Fc region away from the EPO-Rc binding sites on EPO, and (2) keeping one EPO from the other EPO domain, so both EPO domains can interact with EPO-Rc on the erythroid progenitor cells independently. For the present invention, a flexible peptide linker of about 20 or fewer amino acids in length is preferred. It is preferably to use a peptide linker comprising of two or more of the following amino acids: glycine, serine, alanine, and threonine.
The Fc region of human immunoglobulins plays a significant role in immune defense for the elimination of pathogens. Effector functions of IgG are mediated by the Fc region through two major mechanisms: (1) binding to the cell surface Fc receptors (FcγRs) can lead to ingestion of pathogens by phagocytosis or lysis by killer cells via the antibody-dependent cellular cytotoxicity (ADCC) pathway, or (2) binding to the C1q part of the first complement component C1 initiates the complement-dependent cytotoxicity (CDC) pathway, resulting in the lysis of pathogens. Among the four human IgG isotypes, IgG1 and IgG3 are effective in binding to FcγR. The binding affinity of IgG4 to FcγR is an order of magnitude lower than that of IgG1 or IgG3, while binding of IgG2 to FcγR is below detection. Human IgG1 and IgG3 are also effective in binding to C1q and activating the complement cascade. Human IgG2 fixes complement poorly, and IgG4 appears quite deficient in the ability to activate the complement cascade (see, for example, Jefferis et al., Immunol. Rev., 163:59-76, 1998). For therapeutic use in humans, it is essential that when HuEPO-L-Fc binds to EPO-Rc on the surface of the erythroid progenitor cells, the Fc region of the fusion protein will not mediate undesirable effector functions, leading to the lysis or removal of these progenitor cells. Accordingly, the Fc region of HuEPO-L-Fc must be of a non-lytic nature, i. e. the Fc region must be inert in terms of binding to FcγRs and C1q for the triggering of effector functions. It is clear that none of the naturally occurring IgG isotypes is suitable for use to produce the HuEPO-L-Fc fusion protein. To obtain a non-lytic Fc, certain amino acids of the natural Fc region have to be mutated for the attenuation of the effector functions.
By comparing amino acid sequences of human and murine IgG isotypes, a portion of Fc near the N-terminal end of the CH2 domain is implicated to play a role in the binding of IgG Fc to FcγRs. The importance of a motif at positions 234 to 237 has been demonstrated using genetically engineered antibodies (see, for example, Duncan et al., Nature, 332:563-564, 1988). The numbering of the amino acid residues is according to the EU index as described in Kabat et al. (in Sequences of Proteins of Immunological Interest, 5th Edition, United States Department of Health and Human Services, 1991). Among the four human IgG isotypes, IgG1 and IgG3 bind FcγRs the best and share the sequence Leu234-Leu-Gly-Gly237 (only IgG1 is shown in FIG. 1). In IgG4, which binds FcγRs with a lower affinity, this sequence contains a single amino acid substitution, Phe for Leu at position 234. In IgG2, which does not bind FcγRs, there are two substitutions and a deletion leading to Val234-Ala-Gly237 (FIG. 1). To minimize the binding of Fc to FcγR and hence the ADCC activity, Leu235 in IgG4 has been replaced by Ala (see, for example, Hutchins et al., Proc. Natl. Acad. Sci. USA, 92:11980-11984, 1995). IgG1 has been altered in this motif by replacing Glu233-Leu-Leu235 with Pro233-Val-Ala235, which is the sequence from IgG2. This substitution resulted in an IgG1 variant devoid of FcγR-mediated ability to deplete target cells in mice (see, for example, Isaacs et al., J. Immunol., 161:3862-3869, 1998).
A second portion that appears to be important for both FcγR and C1q binding is located near the carboxyl-terminal end of CH2 domain of human IgG (see, for example, Duncan et al., Nature, 332:738-740, 1988). Among the four human IgG isotypes, there is only one site within this portion that shows substitutions: Ser330 and Ser331 in IgG4 replacing Ala330 and Pro331 present in IgG1, IgG2, and IgG3 (FIG. 1). The presence of Ser330 does not affect the binding to FcγR or C1q. The replacement of Pro331 in IgG1 by Ser virtually abolished IgG1 ability to C1q binding, while the replacement of Ser331 by Pro partially restored the complement fixation activity of IgG4 (see, for example, Tao et al., J. Exp. Med., 178:661-667, 1993; Xu et al., J. Biol. Chem., 269:3469-3474, 1994).
We discover that at least three Fc variants (vFc) can be designed for the production of HuEPO-L-vFc fusion proteins (FIG. 1). Human IgG2 Fc does not bind FcγR but showed weak complement activity. An Fcγ2 variant with Pro331Ser mutation should have less complement activity than natural Fcγ2 while remain as a non-binder to FcγR. IgG4 Fc is deficient in activating the complement cascade, and its binding affinity to FcγR is about an order of magnitude lower than that of the most active isotype, IgG1. An Fcγ4 variant with Leu235Ala mutation should exhibit minimal effector functions as compared to the natural Fcγ4. The Fcγ1 variant with Leu234Val, Leu235Ala and Pro331Ser mutations also will exhibit much less effector functions than the natural Fcγ1. These Fc variants are more suitable for the preparation of the EPO fusion proteins than naturally occurring human IgG Fc. It is possible that other replacements can be introduced for the preparation of a non-lytic Fc without compromising the circulating half-life or causing any undesirable conformational changes.
There are many advantages with the present invention. The increased activity and prolonged presence of the HuEPO-L-vFc fusion protein in the serum can lead to lower dosages as well as less frequent injections. Less fluctuations of the drug in serum concentrations also means improved safety and tolerability. Less frequent injections may result in better patient compliance and quality of life. The HuEPO-L-vFc fusion protein containing a non-lytic Fc variant will therefore contribute significantly to the management of anemia caused by conditions including renal failure, cancer chemotherapy, rheumatoid arthritis, AZT treatment for HIV infection, and myelodysplastic syndrome.